what is the meaning of some terms when using `nvcr.io/nvidia/tensorflow:19.03-py2` to convert a tf saved_model?

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

hello, when I use nvcr.io/nvidia/tensorflow:19.03-py2 to convert a tf saved_model, some logs like:

2019-07-11 07:25:46.811638: I tensorflow/contrib/tensorrt/segment/segment.cc:434] Not a TF-TRT candidate, (Op type: Relu), (Op name: resnet_v1_50/conv1/Relu), (Reason: excluded by segmenter option)
2019-07-11 07:25:46.811716: I tensorflow/contrib/tensorrt/segment/segment.cc:434] Not a TF-TRT candidate, (Op type: MaxPool), (Op name: resnet_v1_50/pool1/MaxPool), (Reason: excluded by segmenter option)

why it shows in simple conv1 layer, Relu and MaxPool ops is not supported here? they are the supported ops in trt-tf docs.

2019-07-11 07:08:32.773873: I tensorflow/contrib/tensorrt/segment/segment.cc:696] No device assigned to segment 54
2019-07-11 07:08:32.773897: I tensorflow/contrib/tensorrt/segment/segment.cc:681] Segment 55: parent=resnet_v1_50/block4/unit_2/bottleneck_v1/conv3/BatchNorm/beta/read:
2019-07-11 07:08:32.773917: I tensorflow/contrib/tensorrt/segment/segment.cc:686] Segment 55 has only 0 nodes, dropping
2019-07-11 07:08:32.773943: I tensorflow/contrib/tensorrt/segment/segment.cc:681] Segment 55: parent=resnet_v1_50/block4/unit_2/bottleneck_v1/conv3/BatchNorm/gamma/read:
2019-07-11 07:08:32.773964: I tensorflow/contrib/tensorrt/segment/segment.cc:686] Segment 55 has only 0 nodes, dropping
2019-07-11 07:08:32.773990: I tensorflow/contrib/tensorrt/segment/segment.cc:681] Segment 55: parent=resnet_v1_50/block4/unit_2/bottleneck_v1/conv3/BatchNorm/moving_mean/read:
2019-07-11 07:08:32.774011: I tensorflow/contrib/tensorrt/segment/segment.cc:686] Segment 55 has only 0 nodes, dropping
2019-07-11 07:08:32.774037: I tensorflow/contrib/tensorrt/segment/segment.cc:681] Segment 55: parent=resnet_v1_50/block4/unit_2/bottleneck_v1/conv3/BatchNorm/moving_variance/read:
2019-07-11 07:08:32.774057: I tensorflow/contrib/tensorrt/segment/segment.cc:686] Segment 55 has only 0 nodes, dropping

No device assigned to segment 54 means what?

and logs like:

2019-07-11 07:08:34.901294: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:621] Optimization results for grappler item: tf_graph
2019-07-11 07:08:34.901373: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:623]   constant folding: Graph size after: 828 nodes (-324), 968 edges (-329), time = 1109.96594ms.
2019-07-11 07:08:34.901398: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:623]   layout: Graph size after: 965 nodes (137), 1222 edges (254), time = 193.217ms.
2019-07-11 07:08:34.901506: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:623]   constant folding: Graph size after: 959 nodes (-6), 1222 edges (0), time = 469.334ms.
2019-07-11 07:08:34.901539: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:623]   TensorRTOptimizer: Graph size after: 527 nodes (-432), 782 edges (-440), time = 2563.75ms.

it means what？ and time = 2563.75ms means the time that the TensorRTOptimizer process costs?

2

Hi, can you provide the following details on the platforms you are using?

  1. Linux distro and version
  2. GPU type
  3. Nvidia driver version
    4. CUDA version
    5. CUDNN version
    6. Python version [if using python]
    7. TensorFlow version
    8. TensorRT version
  4. Any source files and models you can provide will help us reproduce your issue and further debug it. You can private message these if you don’t want them to be public.

Thanks,
NVIDIA Enterprise Support